How the Evoa Score works
Evoa Score is an independent 0–100 rating for every AI Agent Skill and MCP server. It exists so you can judge which skill to actually use — not just which has the most GitHub stars. This page discloses exactly how it is computed.
Formula
Evoa Score = Σ ( dimension_score × weight ) × license_factor
where each dimension_score ∈ [0, 100], Σ weights = 100%The 8 dimensions
| Dimension | Weight | What it measures / source |
|---|---|---|
| Task usefulness | 20% | 実タスクにどれだけ役立つか（機能の豊富さ・用途の明確さ）。Source: AIによるcapabilities/use-cases解析 |
| Code quality | 15% | 実装・指示の品質。Source: AIによるSKILL.md/README解析 |
| Maintenance | 15% | リポジトリがどれだけ活発に保守されているか。Source: GitHub 最終push日時の新しさ |
| Documentation | 12% | ドキュメントの充実度・分かりやすさ。Source: README/独自要約の情報量 |
| Security | 15% | 危険・不審な挙動が無いか。Source: AIによるセキュリティレビュー |
| Originality | 10% | ありふれたラッパーではない独自性。Source: AIによる独自性判定 |
| Popularity | 8% | コミュニティの採用度。Source: GitHub Stars/Forks（対数スケール） |
| Compatibility | 5% | 対応AIエージェントの広さ。Source: AIによる対応エージェント判定 |
Principles
- ▸No pay-to-rank. Rankings are determined by this score only and are never influenced by advertising or sponsorships.
- ▸License-aware. ライセンス不明/制限あり（Red）のSkillは総合スコアに0.85倍の補正を適用します。
- ▸Original analysis. Summaries are generated by AI Skills Rank; we never reproduce a skill's source text — we link to the repository.
- ▸Recomputed continuously. Scores update as repositories change (new stars, commits, updated docs).
Weights may be refined over time as we add benchmark data. Browse scored skills →